Drug sensitivity depending on presence of splice altering variants according to SpliceAI predicitons

library(tidyverse)
library(data.table)
library(ggbeeswarm)
library(ggpubr)

rename = dplyr::rename
select = dplyr::select

set wd

Load offtarget splice variations

splice_out = readRDS("spliceai/spliceAI05_Annotated.rds")
#Drug data 
drugs = fread("depmap/drug_sensitivity/primary-screen-replicate-collapsed-logfold-change.csv")
drugs = drugs %>% 
  rename("DepMap_ID" = "V1")

meta = fread("depmap/drug_sensitivity/primary-screen-replicate-treatment-info.csv")

CCLE = fread("depmap/ccle_info_21q1.csv")

|--------------------------------------------------|
|==================================================|
|--------------------------------------------------|
|==================================================|

Join metadata with sensitivity

Tidy up drug data frame

#pivot longer
df_drug = 
  drugs %>% 
  pivot_longer(cols = -DepMap_ID, names_to = "DRUG", values_to = "SENSITIVITY")

#separate information after "::" to new column named "DOSE"
df_drug = 
  df_drug %>%
  separate(DRUG, into = c("DRUG", "DOSE"), sep = "::") %>% 
  mutate(DOSE = as.factor(DOSE))

Expected 2 pieces. Additional pieces discarded in 2708508 rows [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, ...].

#df_drug$DRUG = gsub("::", "", df_drug$DRUG)

head(df_drug)

let’s join with metadata

df_meta = 
  meta %>% 
  rename("DRUG" = "broad_id",
         "DOSE" = "dose") %>% 
  mutate(DOSE = as.factor(DOSE))

df_drug_meta = 
  df_drug %>% 
  left_join(df_meta, by = c("DRUG", "DOSE")) %>% 
  distinct()

head(df_drug_meta)

df_drug_meta = 
  df_drug_meta %>% 
  select(c("DepMap_ID", "DRUG", "DOSE", "SENSITIVITY", "name", "moa", "target")) %>% 
  rename("Hugo_Symbol" = "target") %>% 
  distinct()

head(df_drug_meta)

#split into new rows those drugs that have more than one gene as target (i.e. more than one symbol in column Hugo_Symbol)
df_drug_meta = 
  df_drug_meta %>% 
  mutate(Hugo_Symbol = strsplit(as.character(Hugo_Symbol), ",")) %>% 
  unnest(Hugo_Symbol)

Create new column with isDeleterious information in CCLE We consider deleterious everything that is not a SNP

df_mutations = 
  CCLE %>% 
  select(DepMap_ID, Hugo_Symbol, Variant_Type, isDeleterious, primary_disease)

#head(df_mutations)

#Create new isdel column
df_mutations_del = 
  df_mutations %>% 
  select(-isDeleterious) %>% 
  mutate(isDel = ifelse(Variant_Type == "SNP", F, T))

head(df_mutations_del)

Join with CCLE info

#join
df_drug_mutations = 
  df_drug_meta %>% 
  left_join(df_mutations_del, by = c("DepMap_ID", "Hugo_Symbol")) %>% 
  distinct()

head(df_drug_mutations)

Keep only those drugs who have a gene as target

df_keep = 
  df_drug_mutations %>% 
  filter(!is.na(Hugo_Symbol))

Let’s mark those cell lines with mutations that were predicted by SpliceAI to alter splicing

head(splice_out)

df_splice = 
  splice_out %>% 
  select(DepMap_ID, SYMBOL, exon_intron) %>% 
  rename("Hugo_Symbol" = "SYMBOL") %>% 
  mutate(splice = as.factor(TRUE))

df_splice_drug = 
  df_keep %>% 
  left_join(df_splice, by = c("DepMap_ID", "Hugo_Symbol")) %>% 
  distinct()

Now, let’s keep only the drugs that affect genes that are mutated according to CCLE

only_mutated = 
  df_splice_drug %>% 
  filter(!is.na(Variant_Type))

Filter out IN/DEL mutations

only_mutated = 
  only_mutated %>% 
  filter(isDel == F)

Label NA in “splice” as FALSE

only_mutated = 
  only_mutated %>% 
  mutate(splice = ifelse(is.na(splice), FALSE, splice)) %>% 
  mutate(splice_fac = as.factor(splice))

head(only_mutated)

only_mutated %>% 
  filter(DRUG == "BRD-K35882976-001-04-7")

Combine drug and Symbol

#put SYMBOL between parenthesis
only_mutated$Hugo_Symbol<- gsub("$", ")", only_mutated$Hugo_Symbol)
only_mutated[,7] <- paste0("(", format(unlist(only_mutated[,7])))

head(only_mutated)

combined_mutated = 
  only_mutated %>% 
  unite(drug_symbol, c("name", "Hugo_Symbol"), sep = " ") %>% 
  #drop na values for sensitivity
  drop_na(SENSITIVITY)

Plot

combined_mutated %>% 
  group_by(drug_symbol) %>% 
  mutate(sum_splice_variants = sum(splice)) %>% 
  ungroup() %>%
  filter(sum_splice_variants >= 5) %>% 
  ggplot(aes(y = SENSITIVITY, x = splice_fac, color = splice_fac)) + 
  geom_quasirandom(method = "pseudorandom",alpha = 0.5, size = 0.5) + 
  theme_bw() +
  facet_wrap("drug_symbol") +
  scale_color_manual(labels = c("WT", "splice"), values = c("darkgray","red")) +
  xlab("") +
  ylab("Drug sensitivity (sd from median)") +
  theme(axis.text.x = element_blank(),
        legend.position = "bottom") +
  stat_compare_means(label = "p.format", label.y = -6, label.x = 1, size = 2.5)

ggsave("../../../figures/results/spliceai/drugSensitivity/spliceAI_out_drugSensitivity_filteroutINDELS_DRUG_SYMBOL.png", height = 10, width = 10)

Error in grDevices::dev.off() : 
  QuartzBitmap_Output - unable to open file '../../../figures/results/spliceai/drugSensitivity/spliceAI_out_drugSensitivity_filteroutINDELS_DRUG_SYMBOL.png'

Generate plot facetin per cancer